Abstract: Unmanned vehicle trajectory tracking, a core technology in the field of autonomous driving, provides crucial support for the precise and safe driving of unmanned vehicles. It plays an indispensable role in numerous practical scenarios such as logistics and intelligent transportation. In complex dynamic environments, traditional trajectory tracking methods generally encounter challenges in meeting the application requirements of high precision and reliability owing to their weak dynamic adaptability and insufficient accuracy. To address the challenges associated with weak dynamic adaptability and low accuracy in unmanned vehicle trajectory tracking, this study transformed the unmanned vehicle trajectory tracking problem into a Markov decision process (MDP); designed the state space, action space, and reward function for reinforcement learning; and developed a high-precision trajectory tracking control method for unmanned vehicles based on deep reinforcement learning. The proposed method was implemented and validated using the deep deterministic policy gradient (DDPG) and twin delayed deep deterministic policy gradient (TD3) algorithms, which are suitable for continuous control tasks. First, to enhance the responsiveness of the system to error change rates, differential compensation components for lateral position and heading angle errors were introduced into the state-space design. This enables the agent to perceive the error change trend more accurately during the trajectory tracking process and make control adjustments in advance. Second, a dual-mechanism reward function coordination strategy was proposed to address the difficulty of traditional reward mechanisms in balancing precise rewards and punishments with dynamic adaptability. It is a regionalized reward and punishment mechanism based on a smooth step function. Based on the positional relationship between the unmanned vehicle and the desired trajectory, different reward regions were created, and differentiated rewards and punishments were implemented for the unmanned vehicle in different regions to achieve precise rewards and punishments for the trajectory tracking state. In addition, it is an adaptive weight-reward mechanism based on a Gaussian kernel function. The Gaussian kernel function was used to weigh factors such as errors, allowing the reward function to dynamically adjust the reward weights according to the actual tracking situation and better adapt to different trajectory tracking scenarios. Finally, comprehensive simulation experiments were conducted to validate the effectiveness of the proposed method. Comparative studies with traditional linear quadratic regulator (LQR) control and the original DDPG and TD3 algorithms demonstrated that the proposed approach achieved superior tracking accuracy, smoother control actions, and improved dynamic response under various trajectory tracking scenarios, including straight-line and sinusoidal trajectories. Furthermore, robustness experiments under random noise disturbances indicated that the proposed method could maintain stable control performance and reliable tracking behavior, highlighting its strong robustness and adaptability in complex and uncertain environments. Overall, the results confirmed that the proposed deep reinforcement learning-based trajectory tracking control method effectively balances the dynamic responsiveness and steady-state precision. By jointly improving the state-space representation and reward function design, this method provides a robust and high-precision solution for unmanned vehicle trajectory tracking under complex dynamic conditions.